Surface inspection using texture recognition

This paper presents a novel approach to automating the visual inspection of magnetic disks. In commercial production, quality control is currently achieved by means of functional tests. Visual inspection of the disk surface is required to achieve improved reliability. A novel method for automating visual inspection which relies on texture recognition is presented. The method produces an approximation to the nth order co-occurrence spectrum, which is then processed using pattern recognition techniques. Preliminary results are presented for some frequently occurring disk surface faults. The method is computational efficient and low cost and is shown to be viable in a production environment